Momentum–space structure of relativistic runaway electrons

J. R. Martı́n-Solı́s, J. D. Alvarez, R. Sánchez, and B. Esposito

Citation: Physics of Plasmas (1994-present) 5, 2370 (1998); doi: 10.1063/1.872911

View online: http://dx.doi.org/10.1063/1.872911
View Table of Contents: http://scitation.aip.org/content/aip/journal/pop/5/6?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Investigation of relativistic runaway electrons in electron cyclotron resonance heating discharges on Korea
Superconducting Tokamak Advanced Research
Phys. Plasmas 21, 072509 (2014); 10.1063/1.4890461

Visible imaging and spectroscopy of disruption runaway electrons in DIII-D

Phys. Plasmas 20, 042113 (2013); 10.1063/1.4801738

Note: Measurement of the runaway electrons in the J-TEXT tokamak

Rev. Sci. Instrum. 83, 056108 (2012); 10.1063/1.4721659

A synchrotron radiation diagnostic to observe relativistic runaway electrons in a tokamak plasma

Rev. Sci. Instrum. 72, 466 (2001); 10.1063/1.1318245

Avalanche runaway growth rate from a momentum-space orbit analysis

Phys. Plasmas 6, 2523 (1999); 10.1063/1.873524

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18
 PHYSICS OF PLASMAS VOLUME 5, NUMBER 6 JUNE 1998

Momentum–space structure of relativistic runaway electrons

J. R. Martı́n-Solı́s,a) J. D. Alvarez, and R. Sánchez
Universidad Carlos III de Madrid, Butarque 15, Leganes, 28911-Madrid, Spain
B. Esposito
Associazione Euratom-ENEA CRE, 00044-Frascati, Italy
~Received 24 November 1997; accepted 9 March 1998!
The dynamics of relativistic runaway electrons in tokamak plasmas is analyzed using a test particle
description that includes acceleration in the toroidal electric field, collisions with the plasma
particles, and deceleration due to synchrotron radiation losses. The region of momentum space in
which electron runaway takes place is determined. It is found that relativistic and synchrotron
radiation effects lead to a critical electric field E R .(kT e /m e c 2 )E D , below which no runaways are
generated. In addition, the trajectories of the test electrons in momentum space show a stable
equilibrium point that sets a limit on the energy that the runaway electrons can reach. Analytical
expressions are given for this energy limit as a function of the toroidal electric field and plasma
parameters. The dominant radiative mechanisms limiting the runaway electron energy are identified
in the whole range of electric field values. © 1998 American Institute of Physics.
@S1070-664X~98!02706-2#

I. INTRODUCTION will be on the order of c, the speed of light. Consideration of

relativistic effects yields qualitative differences to the non-
The application of an electric field to a fully ionized relativistic theory because of the significantly different ve-
plasma gives rise to runaway electrons.1 As a result of the locity dependence of Coulomb collisions when v ;c. There-
decrease of the Coulomb collision frequency with increasing fore, the electron distribution function was derived in Ref. 4
energy, electrons with energies higher than some critical en- using a relativistically correct collision integral. The main
ergy are continuously accelerated by the electric field. The results of this analysis were the following: ~1! below the
runaway electrons moving in toroidal paths in the tokamak critical electric field E R 5(kT e /m e c 2 )E D absolutely no run-
speed up very quickly until the synchrotron radiation loss aways are produced; ~2! for electric fields E i .E R , the criti-
due to the curved path, which increases with the electron cal velocity at which electrons run away is modified by rela-
energy, balances the energy gain in the electric field. In large tivistic effects. In terms of the electron momentum, runaway
tokamaks, runaway electrons confined for a sufficiently long generation occurs for p i .m e c/ AD21, with D5E i /E R .
time can gain enough energy to cause serious damage to the The motion of relativistic test runaway electrons in momen-
confining structures. This is particularly important for dis- tum space, under electric field acceleration and collisions
ruption generated runaway electrons.2 with the plasma particles, was addressed by Fussmann in
In the past, phase-space analysis of the relaxation equa- Ref. 6.
tions describing nonrelativistic test electrons was used to in- In this paper we investigate the dynamics of relativistic
vestigate the conditions for runaway.3 The electron dynamics test electrons in momentum space, including the acceleration
was analyzed, taking into account the acceleration in the in the toroidal electric field, collisions with the plasma par-
electric field and the collisions with the plasma particles. The ticles, and deceleration due to synchrotron radiation losses.
condition v 2i >(21Z eff)1/2(E D /E i )•(kT e /m e ) for runaway As in Refs. 3 and 6, the electron motion shows a saddle point
generation was inferred, and it was shown that, in general, in momentun space around which the electron trajectories
runaway in velocity space only occurs for electrons that are separate, and the runaway region lies just outside the sepa-
outside one of the separatrices of the relaxation equations. ratrix of motion passing through the saddle point. The escap-
The results, based on single-particle dynamics, were cor- ing electron trajectories move toward a stable equilibrium
roborated by numerical integration of the two-dimensional point in which the energy gain in the toroidal electric field is
Fokker–Planck equation. balanced by collisional and radiation losses, setting a limit
Connor and Hastie4 showed that, if the electric field is on the energy that the runaway electrons can reach. In Sec. II
sufficiently weak, E i /E D ;kT e /m e c 2 , relativistic effects be- the test electron relaxation equations are presented together
come important for runaway generation. The condition with a phase-space analysis of the equations. In Sec. III the
above can be met in present large tokamaks and the proposed main results related to runaway generation ~critical momen-
International Thermonuclear Experimental Reactor ~ITER!.5 tum and critical electric field for runaway generation! are
In this case, the critical velocity v c for runaway generation described and compared to those in Refs. 3, 4, and 6. In Sec.
IV the runaway energy limit is calculated, for given plasma
a!
Electronic mail: solis@fis.uc3m.es parameters, as a function of the toroidal electric field. It is

1070-664X/98/5(6)/2370/8/$15.00 2370 © 1998 American Institute of Physics

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18
 Phys. Plasmas, Vol. 5, No. 6, June 1998 Martı́n-Solı́s et al. 2371

shown that, for low values of the electric field, the main
mechanism limiting the runaway energy are the synchrotron
losses due to the gyromotion of the runaways having a finite
pitch angle, while for high electric field values the synchro-
tron losses are dominated by the electron motion on toroidal
paths. The conclusions are summarized in Sec. V.

II. TEST ELECTRON RELAXATION EQUATIONS

The dynamics of a relativistic electron in a tokamak
plasma is described using the test particle equations:
dp i n c e 4 ln Lm e pi pi
5eE i 2 g ~ Z eff111 g ! 3 2F S ,
dt 4pe0 2
p p
~1!
dp p i n e e 4 ln Lm e g 2
5eE i 2 2F S , ~2! FIG. 1. The phase-space plot of Eqs. ~5! and ~7! for D54.4, a54, F gy
dt p 4 p e 20 p2 50.65, and F gc52.331028 . Here, P 1 is the saddle point, P 2 is the stable
where p i is the electron momentum parallel to the magnetic focus, S r and S a are the separatrices, and q c the critical ~normalized! mo-
mentum for runaway generation.
field ~approximately in the toroidal direction!, p is the total
electron momentum, and g is the relativistic gamma factor,
g 2 511p 2 /(m e c 2 ) 2 ; e is the absolute value of the electron We rewrite Eqs. ~1! and ~2! in normalized form,

S D SD
charge, m e is the electron mass, E i is the toroidal electric
field, Z eff is the effective ion charge, and ln L is the Coulomb dq i qi q'2 v 3
qi
5D2 g ~ a 1 g ! 3 2 F gc1F gy 4 g 4 ,
logarithm. dt q q c q
The first term in the equations ~1! and ~2! is the accel- ~5!
eration due the toroidal electric field, and the second term
includes the effect of the collisions with the plasma
particles.7 The effects of the synchrotron radiation losses are
dq
dt
qi g2
S
q'2
5D 2 2 2 F gc1F gy 4 g 4
q q q
v
c
D SD 3
, ~6!

described by means of a decelerating force:8 where q5p/m e c and t 5 n r t, with nr

SD K L 3 5n e e 4 ln L/4p e 20 m 2e c 3 ; D5E i /E R is the normalized elec-

2 v 1
F S5 r em ec 2 g4 , ~3! tric field, with E R 5(kT e /m e c 2 )E D , where E D
3 c R2 5n e e 3 ln L/4p e 20 kT e is the Dreicer field; a 511Z eff , and
with the radius of curvature averaged over one gyrorotation,9 F gc , F gy are parameters describing the two contributions to

K L F
the radiation losses ~the guiding center motion and the elec-
1 1 tron gyromotion, respectively!, given by F gc5F gy(m e c/
2 5 2 cos u 1cos u sin u
6 2 4
R R0 eB 0 R 0 ) 2 , F gy52 e 0 B 20 /3n e , ln Lme (B 0 is the toroidal mag-

S DG
netic field!.
5 r 2g 1
1cos4 u sin2 u 1 1 sin4 u ~4! From ~5! and ~6! we get

F
2 8R 20 r 2g
dq' 1 q 2i
(R 0 is the plasma major radius, u the pitch angle, and r g is q' 5 g~ a 1 g ! 22 g 2
dt q q
the electron gyroradius!; r e 5e 2 /(4 p e 0 m e c 2 ) is the classical
electron radius and v is the electron velocity, v /c5( g 2
21) 1/2/ g .
Note that the orbit of an electron in a tokamak is com-
S
2 F gc1F gy
q'2
q4
D SD G
g4
v
c
3
q'2 . ~7!

posed of two parts: the guiding center motion ~along the The system of equations ~5! and ~7! have in (q i ,q'2 )
magnetic field lines! and the gyromotion around the field space two singular points ~a saddle point, P 1 , and a stable
lines, both contributing to the radius of curvature of the elec- focus, P 2 ) as it is shown in the phase-space plot of Fig. 1.
tron. The first term in ~4! is mainly determined by the guid- An analytical relation between the singular points and the
ing center motion and the second term by the electron gyro- normalized electric field can be obtained considering that P 1
motion. Since r g !R 0 , the second term is dominant for finite and P 2 lie at the intersection of the two contours q̇ i 50 and
pitch angles ~and so, the radius of curvature is determined by q̇' 50 ~or q̇50). Thus, from q̇50,
the electron gyromotion!; only when u is close to zero, the g 2s
contribution due to the guiding center motion is non- D5
cos u s ~ g 2s 21 !
negligible. Thus, we can use for ^ 1/R 2 & the simplified rela-
tion
S ~ g 2s 21 ! 3/2 ~ g 2s 21 ! 5/2
D
K L
3 11F gy sin2 u s 1F gc , ~8!
1 1 sin u 4
gs gs
2 . 21 .
R R0 r 2g with cos us , from q̇' 50, given by
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18
 2372 Phys. Plasmas, Vol. 5, No. 6, June 1998 Martı́n-Solı́s et al.

point in momentum space ~the saddle point P 1 ) and the re-

lation between this point and the normalized electric field
can be readily obtained from q̇ i 50, q̇' 50 and using F S
50 in the single-particle equations, yielding

g 3/2
s ~a1gs!
1/2
D5 ~10!
g 2s 21
( g s is the relativistic gamma factor at the singular point!.
The relation ~10! is also shown in Fig. 2 for comparison. In
this case, there is only one branch ~the one corresponding to
the saddle point! extending to infinity in g s .

FIG. 2. The normalized electric field versus g s at the singular points: full
line, taking into account the radiation losses; dashed lined, neglecting the
power radiated by the electron. III. RUNAWAY GENERATION
A. Critical electric field
The nonrelativistic theory of a plasma in an electric field
F gc~ g 2s 21 ! a1gs
cos u s 511
2
1 E i predicts that there will always be runaway electrons. This
2F gy 2F gy~ g 2s 21 ! 3/2 results from the fact that the momentum loss due to colli-

H F
3 12
4 a F gy~ g 2s 21 ! 3/2
~ a1gs!2
sions decreases with increasing energy, d p/dt; p 22 , so that
for any value of the electric field, above some critical energy,
the electrons will be continuously accelerated. However, if

S
1 11
F gc~ g 2s 21 ! 5/2
a1gs
DG J
2 1/2
. ~9!
relativistic effects are considered, as momentum increases,
v →c, d p/dt; v 22 , and, unlike the nonrelativistic case, the
collision loss is no longer reduced, implying that for weak
In these equations, g s is the relativistic gamma factor at enough electric fields no electrons will run away. The de-
the singular points @to which we will often refer as the elec- tailed analysis done by Connor and Hastie4 allows to define a
tron energy, given the straightforward relation between g and critical electric field, E R 5(kT e /m e c 2 )E D ~in normalized
the electron energy, E5( g 21)m e c 2 #, and u s the pitch form D R 51), below which no electrons run away.
angle. Figure 2 shows, for given plasma parameters, the re- The work done in Ref. 4 considered the acceleration in
lation ~8! between D and the energy g s of the singular the toroidal electric field and the collisions of the relativistic
points. For each value of the normalized electric field D, electrons with the bulk plasma. In fact, the result D R 51 can
there are two values of g s corresponding to the singular be easily recovered from the analysis of the test particle
points P 1 and P 2 ; branch I in Fig. 2 gives g s at the saddle equations when F S 50. If the electron radiation is not con-
point P 1 , while branch II gives the electron gamma value at sidered, the relation between the normalized electric field D
the stable focus P 2 . The two branches are separated by the and the saddle point energy is given by ~10!: as shown in
minimum of D vs g s . As it will be explained in detail in the Fig. 2, D is a decreasing function of g s , and taking the limit
following sections, the two singular points have a well- g s →`, we get D→1, so that for D,1 there is no singular
defined physical meaning: the saddle point ~branch I! gives point in momentum space and no runaway electrons are gen-
the critical energy for runaway generation, and the stable erated. This can be more easily understood when considering
focus ~branch II! the energy limit for the generated runaway that, as will be explained in the next section, the saddle point
electrons. energy constitutes an estimate of the critical energy for run-
The limiting particle trajectories passing through P 1 and away generation: if the electric field decreases, the critical
P 2 are the separatrices S r and S a . The separatrix S r divides energy for runaway generation ( g s ) increases and therefore,
the phase space (q i ,q'2 ) in two regions, the region outside S r in the limit g s →`, we get the critical D R below which ab-
constituting the runaway region: all the electrons initially solutely no runaways are generated.
outside S r will eventually move along S a toward the stable When the radiation losses are taken into account, the
focus P 2 . Electrons initially inside S r will collapse into the relation between D and the singular points is given by ~8!.
origin ~i.e., they do not run away!. This relation ~see Fig. 2! shows a local minimum that sepa-
It is convenient to remark that the stable point P 2 is a rates the saddle points ~branch I! from the stable points
consequence of including the radiation losses in the test ~branch II!. As it was stated in Sec. II, for each value of the
equations. The runaway electrons are not continuously accel- electric field, branch I gives the critical energy g c for run-
erated by the toroidal electric field but they reach a maxi- away generation and branch II the runaway energy limit g l .
mum energy when the power radiated by the electron equals When D decreases, less energy is absorbed from the electric
the energy gain in the electric field. This energy balance field so that g c increases ~it is more difficult to generate
takes place at P 2 , which sets a limit on the energy that the runaway electrons! and the energy limit g l decreases, until
runaway electrons can reach. If the radiation losses are not both of them coalesce for a given value of D ~the minimum
considered, the relaxation equations only show one singular of D versus energy! below which no electrons will run away.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18
 Phys. Plasmas, Vol. 5, No. 6, June 1998 Martı́n-Solı́s et al. 2373

FIG. 4. A comparison between the critical ~normalized! momentum for

runaway generation calculated numerically and the ~normalized! momentum
FIG. 3. Critical electric field D R for runaway generation versus F gy ; g R at the saddle point, versus the normalized electric field.
gives the critical energy for runaway generation at D R .

come the energy losses. The highest critical momentum is

Thus, the minimum electric field D R for runaway generation achieved at the critical electric field D R .
can be deduced from the condition dD/d g s 50, with D In Ref. 4, the critical momentum q c was calculated from
5D( g s ) given by ~8!. a kinetic analysis of the electron distribution function in
The critical electric field D R calculated from the above which the radiation losses were neglected. In such a case,
condition is plotted in Fig. 3 versus F gy , the dimensionless F gy50, and, following a single-particle analysis, the relation
parameter defined in Sec. II characterizing the radiation term between D and the critical g c , estimated from the saddle
strength. In the nonradiative limit (F gy→0), D R →1, recov- point energy g s , is given by ~10!. Taking the limit consid-
ering the result given in Ref. 4. However, when F gy.0, the ered in Ref. 4, E i /E D;kT e /m e c 2 ~i.e., D;1), then g c @1
momentum and energy losses increase due to the radiated and Eq. ~10! can be approximated by D; g 2c /( g 2c 21), yield-
power, leading to more restrictive conditions on runaway ing g c ; A(D21)/D and q c ;1/AD21, which is the result
generation so that D R .1. Furthermore, D R increases with given in Ref. 4.
F gy and the critical energy for runaway generation keeps In Fig. 5, a comparison between the critical momentum
finite, even if D→D R . The critical energy for runaway gen- obtained from the single-particle analysis for a case with
eration when D→D R is also plotted in Fig. 3 as a function of F gy50 and a case with F gy.0 is shown. The radiation
F gy : this energy ~named g R in the figure! decreases with losses lead to higher values of q c when D is close to the
increasing F gy and, when F gy→0, goes to infinity. critical field D R ; on the other hand, the deceleration due the
synchrotron losses imposes severe limits on the values that
q c can reach, even when D→D R , in contrast to the nonra-
B. Critical momentum diative case, in which q c →` when D→1. This can reduce
greatly the relativistic effects on the critical momentum q c
In Sec. II, it has been shown that the separatrix S r deter- for reasonable values of F gy .
mines the region in momentum space in which runaway gen- A characteristic feature of S r seen in Fig. 1 is its limited
eration takes place. However, it is not easy to determine extent on both sides of the q i axis, indicating the presence of
explicitly S r from Eqs. ~5! and ~7!. As pointed out in Ref. 3, runaways originating in a direction opposite to the force
since most of the electrons must be concentrated in the di- caused by the electric field (q i ,0), and called ‘‘backward’’
rection associated to the electric field force, the criterion for
runaway generation can be reduced to the condition q i
.q c , where q c is the intersection of the separatrix S r with
the positive q i axis. Therefore, q c can be defined as the criti-
cal ~normalized! momentum for runaway generation.
The value of q c can be obtained numerically or esti-
mated following the considerations given in Ref. 3: the criti-
cal energy and momentum ( g c and q c ) for runaway genera-
tion can be approximated by those at the saddle point P 1 .
Thus, g c will be given by branch I in Eq. ~8! and q c 5( g 2c
21) 1/2. A comparison between the numerically calculated q c
and the electron momentum at P 1 is shown in Fig. 4 as a
function of the normalized electric field. Although P 1 gives
an overestimate of q c , the approximation is reasonably
FIG. 5. A comparison between the critical momentum q c calculated for
good. The critical momentum q c increases when the electric F gy50 ~no rad., i.e., no radiation considered! and for F gy50.65 ~rad., i.e.,
field decreases: the electrons absorb less energy from the radiation losses taken into account!, versus the normalized electric field. The
electric field, and more initial momentum is required to over- critical electric field D R when F gy50.65 is also indicated.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18
 2374 Phys. Plasmas, Vol. 5, No. 6, June 1998 Martı́n-Solı́s et al.

FIG. 7. The runaway energy limit g l versus the normalized electric field
@trace ~a!#. The energy limits calculated assuming that the radiation losses
are dominated by the electron gyromotion @trace ~b!#, and by the motion on
toroidal paths @trace ~c!# are also shown.

portant due to the high perpendicular momentum gained

through collisions, leading to finite pitch angles and impor-
tant radiation losses associated to the electron gyromotion.
The differences between q c f and q cb increase sharply close
to the critical electric field D R , when the radiation losses are
FIG. 6. ~a! Critical momentum q c f for runaway generation in the forward larger. The role played by the radiation losses in determining
direction and critical momentum q cb for backward runaway generation ver- q cb is illustrated in Fig. 6~b!, in which the backward critical
sus the normalized electric field; ~b! critical momentum q cb for backward momenta calculated for F gy.0 and F gy50 are compared:
runaway generation for F gy50 ~no rad.! and for F gy50.65 ~rad.! versus the the radiation dominates when getting close to D R .
normalized electric field. The critical electric field D R for F gy50.65 is in-
dicated.

IV. ENERGY LIMITS

10
runaway electrons. These electrons initially travel counter
One of the main features of the single-particle equations
to the force exerted by the electric field: the field decelerates
introduced in Sec. II is the presence of a stable focus P 2 in
the electrons and collisional effects become important. How-
momentum space toward which all the electron trajectories
ever, if q' is still large enough when q i 50, the electrons are
outside S r move to. The energy g l corresponding to this
accelerated by the electric field and can run away just like
singular point gives the energy limit for the generated run-
the forward runaway electrons. As pointed out by Fisch and
away electrons: the escaping electrons reach a maximum en-
Karney,11 the backward runaways could be important during
ergy at P 2 , where the energy gain in the toroidal electric
lower-hybrid current ramp-up, draining out energy of the
field is compensated by collisional and radiation losses. The
electric field, and leading to a degradation of the ramp-up
relation between the energy limit g l and the normalized elec-
efficiency.
tric field is given by the branch II of Eq. ~8!, and it is shown
We can define, as in the case of the forward runaway
in Fig. 7.
electrons, the critical momentum q cb for backward runaway
As explained in Sec. II, the power radiated by the elec-
generation as the minimum momentum for which electrons
tron is the result of the electron motion on toroidal paths ~the
initially moving counter to the electric field become run-
guiding center motion! and the electron gyromotion around
aways. This momentum q cb is given by the intersection of S r
the field lines. In order to assess the role played by each of
with the negative q i axis and is the relevant quantity during
these mechanisms in determining the runaway electron en-
lower-hybrid ramp-up. The comparison of q cb with the pre-
ergy, the runaway energy limit has been calculated in the two
viously defined critical momentum q c ~which we will denote
following cases.
now as q c f to stress that it is referred to forward runaway
~i! The radiation losses are assumed to be dominated by
electrons! can serve as a quantitative measurement of the
the electron gyromotion. In this case, the runaway energy
differences between forward and backward runaway genera-
limit is determined from Eqs. ~6! and ~7!, neglecting the term
tion. These differences are illustrated in Fig. 6~a!, in which
in F gc and using that in the equilibrium q̇5q̇' 50, giving
q c f and q cb are plotted as function of the normalized electric
field. It is clear that to generate backward runaway electrons
is always more difficult than to generate runaway electrons
in the forward direction: the electric field drains energy from
D5
g 2gy
cos u gy~g2gy21)
11F gyS~g2gy21!3/2 2
ggy
sin u gy , D
~11!
the electrons and the collisional losses increase. This effect is
enhanced by synchrotron deceleration, which becomes im- with
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18
 Phys. Plasmas, Vol. 5, No. 6, June 1998 Martı́n-Solı́s et al. 2375

a1ggy A5 @ 27a 2 1 ~ 729a 4 16912b 3 ! 1/2# 1/3,

cos2 u gy511 ~17!
2F gy~g2gy21) 3/2 4•2 1/3b A
1

F S DG
B52 .
4 a F gy~g2gy21!3/2 1/2 A 3•2 1/3
3 12 11 . ~12! The analytical expression for g l takes a particularly
~a1ggy) 2
simple form in the two limiting cases discussed before. In the
In these relations, the subscript gy has been used to remark first case, when the radiation losses are assumed to be domi-
that only the contribution associated to the electron gyromo- nated by the electron gyromotion @a F gyg l /D@F gcg 4l in
tion is considered in the radiation terrn. ~14!#, we have
~ii! The radiation losses in this case are assumed to be
D ~ D21 !
dominated by the electron motion on toroidal paths, ^ 1/R 2 & g l. . ~18!
;1/R 20 , so that the energy limit is estimated neglecting in the a F gy
test electron equations the term in F gy . Thus, using again The condition of validity of this result, a F gyg l /D@F gcg 4l ,
q̇5q̇' 50 in Eqs. ~6! and ~7!, we have can now be rewritten using ~18! as D 4 (D21) 3

D5
g 2gc
g2gc21
11F gcS~g2gc21!5/2
ggc
D !( a F gy) 4 /F gc , showing that, as it was discussed before, the
radiation associated to the electron gyromotion will be domi-
nant at low values of the electric field. Note that in ~18!, as a

3 11 S a
ggc1F gc~g2gc21) 5/2 D 1/2
. ~13!
consequence of the pitch angle scattering due to collisions,
g l decreases when Z eff and n e increase.
In the second limiting case, the radiation is assumed to
be dominated by the guiding center motion ( a F gyg l /
In Fig. 7, the energies g gy and g gc obtained from the above
D!F gcg 4l ) and then, from ~14!,
expressions are plotted together with g l calculated using the
general relation ~8!. From this figure, it is easy to infer that,
for low values of the electric field, the runaway energy limit g l. S D
D21
F gc
1/4
. ~19!
is dominated by the radiation associated to the electron gy-
romotion and g l can be derived from ~11! and ~12! since, In a similar way to the first case, we can rewrite the range of
when the electric field is low, the collisions of runaways with applicability of this result ( a F gyg l /D!F gcg 4l ), using ~19!,
plasma particles are effective enough to increase the pitch as D(D21) 3/4@ a F gy /F 1/4
gc . Therefore, the relation ~19! for
angle, and therefore the radiation losses associated to the g l will be valid for high enough electric field values.
electron gyromotion. However, for high electric fields, the
effect of the collisions with the bulk plasma is negligible and V. CONCLUSIONS
the pitch angle is close to zero, so that the radius of curvature
is almost equal to the major radius of the tokamak and the In this paper, the trajectories in momentum space of rela-
energy limit is well approximated by ~13!. tivistic test electrons have been analyzed, taking into account
Although the relation ~8! between D and g l is valid for the acceleration in the toroidal electric field, the collisions
any value of the electric field, it is often convenient to have with the plasma particles, and the deceleration due to syn-
explicit expressions giving g l as a function of D. An explicit chrotron radiation losses. The single-particle relaxation equa-
expression for g l can be derived under the assumptions g 2l tions for q i and q'2 are characterized in (q i ,q'2 ) space by the
@1 and g l @ a /2D, conditions that are usually verified for presence of two singular points: a saddle point around which
most of the electric field values ~see Fig. 7 as an example!. the trajectories separate ~collapsing into the origin or moving
Under these conditions it is not difficult to show ~see the out to larger energies! and a stable focus toward which all
Appendix! that cos ul.1 and sin2 ul.a/Dgl , where u l is the the escaping trajectories move. The relation between the
pitch angle at P 2 . Then, from ~8! and using that g 2l 21 ~normalized! electric field and the singular points, D
. g 2l when g 2l @1, we get 5D( g s ), is given by Eqs. ~8! and ~9!.
Previous investigations4 showed that, due to relativistics
a F gy effects, if D→1 the critical momentum for runaway genera-
D.11 g l 1F gcg 4l . ~14!
D tion recedes to infinity, i.e., there are no runaway electrons
produced if D,D R 51. The inclusion of the synchrotron ra-
This relation constitutes an algebraic quartic equation, diation into the runaway dynamics, increasing the energy and
momentum losses, and precluding the runaway generation at
g 4l 1 ag l 2b50 ~15!
large energies, leads to a critical electric field D R greater
@a[ a F gy /DF gc , b[(D21)/F gc# with a real positive solu- than unity that increases with the radiation strength ~de-
tion, scribed by the parameter F gy defined in Sec. II!. It has been
found that, within the single-particle description, D R is the

g l5
1
2F 2a
2B 1/21 2B1 1/2
B S DG1/2
, ~16!
electric field for which the two singular points in momentum
space ~the saddle point and the stable focus! coalesce. This
value corresponds to the minimum of D vs g s and can be
and found from the condition dD/d g s 50 in ~8!.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18
 2376 Phys. Plasmas, Vol. 5, No. 6, June 1998 Martı́n-Solı́s et al.

In general, as explained in Ref. 3, runaway in momen- ripple interaction. In a tokamak, the ripple strength is larger
tum space will occur for those electrons that are outside the for low toroidal harmonic numbers and decays rapidly with
separatrix S r passing through the saddle point. However, radius from the plasma edge. Therefore, it is expected that
since most of the electrons must be concentrated in the di- the most important effects will be found at the edge plasma
rection of the electric field force, a simplified condition q i and for low values of the toroidal harmonic number. Experi-
.q c for runaway generation can be established with q c , the mental evidence for efficient runaway–ripple interaction has
critical momentum for runaway generation, given by the in- been reported in several devices,9,12,13 and it has been pre-
tersection of the separatrix S r with the positive q i axis. It has dicted that a resonance between the electron gyrofrequency
been shown that a reasonable estimate of q c is given by the and the fundamental ripple frequency could create an upper
~normalized! electron momentum at the saddle point, ob- bound on runaway energy during a plasma disruption in
tained from Eqs. ~8! and ~9!. The radiation losses increase large tokamaks like the Joint European Torus ~JET!14 or
the value of q c in comparison with a nonradiative theory, the ITER.15 An analysis of the dynamics of the interaction be-
largest effect taking place for electric fields close to the criti- tween the toroidal field ripple and the runaway motion using
cal electric field D R ; moreover, due to the synchrotron ra- the test particle description here presented has already been
diation, the critical momentum q c always keeps finite, even addressed in a previous work,16 and will be the subject of a
when D→D R . The effect of the radiation is larger on the future publication.
generation of runaways in the backward direction due to the ~b! An outward drift of the electron orbit when its en-
high perpendicular momentum gained through collisions by ergy increases: the drift orbit of a runaway electron gener-
electrons traveling counter to the electric field, which will ated at an initial radius r c , measured from the magnetic axis,
enhance the radiation stemming from the electron gyromo- will be shifted to the low magnetic field side of the flux
tion; again, the most noticeable effects will occur when D surface until it intersects a physical barrier ~a limiter or first
→D R . wall! at a radius r l . A simple estimate ~assuming a flat cur-
The generated runaway electrons are not continuously rent profile! of the energy g for which this occurs is given17
accelerated but they reach a maximum energy when the en- by

FS S D D G
ergy gain in the toroidal electric field is balanced by the rc Ip 2 1/2
collisional and radiation losses. This balance takes place in g5 2R 0 12 11 , ~20!
momentum space at the stable focus. Thus, the energy limit r l 17 000r l
for the runaway electrons will be equal to the electron energy where I p is the plasma current.
at the stable focus, which, for given electric field and plasma ~c! Loss mechanisms that can deplete the population of
parameters, can be determined from the relations ~8! and ~9!. runaways, such as turbulent diffusion processes,18 having an
An explicit expression for the energy limit g l , valid for a influence on the limiting electron energy. These losses, and
wide range of electric field values, has been given. It has their effect on the runaway energy, are usually difficult to
been shown that at low electric fields, the collisions with the quantify.
bulk plasma are effective enough to increase the pitch angle,
and thus the synchrotron losses, to values finally leading to
the energy blocking of runaways; the energy limit in this ACKNOWLEDGMENTS
case increases quadratically with the electric field and de-
creases when the effective ion charge Z eff and the electron J. R. Martı́n-Solı́s wishes to thank P. J. Hernando for
density n e increase. At high electric field values, the radia- helpful discussions and valuable suggestions on the phase-
tion is dominated by the electron motion on toroidal paths space analysis.
and the energy limit g l increases with the electric field as This work was done under financial support from Direc-
g l ;E 1/4
i . ción General de Investigación Cientı́fica y Técnica ~DGI-
There are a number of other mechanisms limiting the CYT! Project No. PB93-0231-C02-01.
energy that runaway electrons can reach. Thus, an upper
boundary on the runaway energy can also appear due to the
following. APPENDIX: PITCH ANGLE VALUE AT THE STABLE
~a! A resonance between the electron gyromotion and FOCUS
the harmonics of the toroidal field ripple, which increases the
energy perpendicular to the magnetic field and therefore the The value of the pitch angle u l at the stable focus P 2 ,
power radiated by the electrons.12 The interaction with the under the assumptions g 2l @1 and g l @ a /2D, can be derived
nth harmonic of the toroidal field ripple takes place at an as follows. At the singular point, q̇ i 50 and q̇50, so that
energy g n .eB 0 R 0 /(nN c m e c), 12 where N c is the number of Eqs. ~5! and ~6! will be written as
toroidal field coils and n is the toroidal harmonic number: if
g n , g l ~the radiation limit considered in Sec. IV! the ripple
resonance can set an additional barrier to the increase in the
05D2 g ~ a 1 g !
qi
q3
S
2 F gc1F
q
D SD
q'2 4 v
gy 4 g
c
3
qi
q
, ~A1!

runaway energy. On the other hand, the efficiency of the

ripple mechanism to block the runaway energy is strongly
dependent on the ripple amplitude: at a given electric field, a
05D
qi g2
q q
S q'2
2 2 2 F gc1F gy 4 g 4
q
v
c
D SD 3
, ~A2!

large enough ripple amplitude is necessary for an efficient where all the dynamical quantities are referred to as the sin-
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18
 Phys. Plasmas, Vol. 5, No. 6, June 1998 Martı́n-Solı́s et al. 2377

gular point. Multiplying ~A2! by q i /q and substracting it 1

H. Dreicer, Phys. Rev. 115, 238 ~1959!.
from ~A1!, the radiation term can be eliminated: 2
R. Jaspers, N. J. Lopes-Cardozo, and F. C. Schüller, Nucl. Fusion 36, 367

S D
~1996!.
q 2i qi 3
V. Fuchs, R. A. Cairns, C. N. Lashmore-Davies, and M. M. Shoucri, Phys.
05D 12 2 ag , ~A3! Fluids 29, 2931 ~1986!.
q2 q3 4
J. W. Connor and R. J. Hastie, Nucl. Fusion 15, 415 ~1975!.
5
ITER EDA Agreement and Protocol 2 ~International Atomic Energy
and using cos ul5qi /q, we obtain Agency, Vienna, 1994!.
a gl
6
G. Fussmann, Nucl. Fusion 19, 327 ~1979!.
cos2 u l 1 cos u l 2150, ~A4! 7
D. Mosher, Phys. Fluids 18, 846 ~1975!.
D g 2l 21 8
J. Schwinger, Phys. Rev. 75, 1912 ~1949!.
9
B. Kurzan, K.-H. Steuer, and G. Fussman, Phys. Rev. Lett. 75, 4626
which is a quadratic equation in cos ul , whose solution is ~1995!.

F S DG
10
N. J. Fisch, Rev. Mod. Phys. 59, 1775 ~1987!.
a gl a gl 2 1/2
11
N. J. Fisch and C. F. F. Karney, Phys. Rev. Lett. 54, 897 ~1985!.
cos u l 52 1 11 . ~A5!
2D g 2l 21 2D g 2l 21 12
L. Laurent and J. M. Rax, Europhys. Lett. 11, 219 ~1990!.
13
R. J. E. Jaspers, Ph.D. thesis, Technical University Eindhoven, 1995.
14
Using the conditions g 2l @1 ( g 2l 21. g 2l ) and g l @ a /2D, it P. H. Rebut, R. J. Bickerton, and B. E. Keen, Nucl. Fusion 25, 1011
~1985!.
is easy to get 15
A. J. Russo, Nucl. Fusion 31, 117 ~1991!.
a
F S DG
a 2 1/2 16
R. Martin-Solis, J. D. Alvarez, R. Sanchez, and B. Esposito, in Controlled
cos u l .2 1 11 .1, ~A6! Fusion and Plasma Physics, Proceedings of the 24th EPS Conference,
2D g l 2D g l Berchtesgaden, 1997, edited by M. Schittenhelm, R. Bartiromo, and F.
Wagner ~European Physical Society, Petit-Lancy, 1997!, Vol. 21A, Part II,
and from ~A4!, p. 865.
17
H. Knoepfel and D. A. Spong, Nucl. Fusion 19, 785 ~1979!.
a gl a
sin2 u l 5 cos u l .
18
. ~A7! O. J. Kwon, P. H. Diamond, F. Wagner, G. Fussmann, the ASDEX team,
D g 2l 21 Dgl and the NI Teams, Nucl. Fusion 28, 1931 ~1988!.

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
128.138.73.68 On: Sun, 21 Dec 2014 16:09:18